--- trunk/all2xml.pl 2003/12/15 00:12:16 196
+++ branches/biomed/all2xml.pl 2004/06/02 16:13:57 335
@@ -5,7 +5,6 @@
use Getopt::Std;
use Data::Dumper;
use XML::Simple;
-use Text::Unaccent 1.02; # 1.01 won't compile on my platform,
use Text::Iconv;
use Config::IniFiles;
use Encode;
@@ -17,6 +16,7 @@
my $config_file = $0;
$config_file =~ s/\.pl$/.conf/;
+$config_file = $ARGV[0] if ($ARGV[0] && -f $ARGV[0]);
die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file);
my $config;
@@ -151,6 +151,58 @@
}
+ # init variables for different types
+ sub init_visible_type($) {
+ my $type = shift;
+
+ # swish, swish_exact, display, index, index_lookup
+ # swish and display defaults
+ my ($s,$se,$d,$i,$il) = (1,0,1,0,0);
+ if (lc($type) eq "display") {
+ $s = 0;
+ } elsif (lc($type) eq "swish") {
+ $d = 0;
+ } elsif (lc($type) eq "index") {
+ ($s,$se,$d,$i) = (0,1,0,1);
+ } elsif (lc($type) eq "swish_exact") {
+ ($s,$se,$d,$i) = (0,1,0,0);
+ } elsif (lc($type) =~ /^lookup/) {
+ ($s,$se,$d,$i,$il) = (0,1,0,0,1);
+ } elsif ($type) {
+ print STDERR "WARNING: unknown type: $type\n";
+ }
+ return ($s,$se,$d,$i,$il);
+ }
+
+
+ # convert
+ #
+ #
+ # ,
+ # 200a
+ #
+ #
+ # to
+ #
+ # 200a
+ #
+ # but without loosing spaces in delimiter (becasue
+ # new XML::Simple strips spaces in attribute values
+ # as defined in XML specification)
+ #
+ sub unroll_x($) {
+ my $x = shift;
+
+ if (defined $x->{value}) {
+ my ($v,$d) = ($x->{value}->{content}, $x->{delimiter}->{content});
+ delete $x->{value};
+ delete $x->{delimiter};
+ $x->{content} = $v;
+ $x->{delimiter} = $d;
+ }
+ return $x;
+ }
+
# begin real work: go field by field
foreach my $field (@sorted_tags) {
@@ -184,23 +236,18 @@
foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
+ $x = unroll_x($x);
+
my $format = x($x->{content});
my $delimiter = x($x->{delimiter}) || ' ';
my $repeat_off = 0; # init repeatable offset
- # swish, swish_exact, display, index, index_lookup
- # swish and display defaults
- my ($s,$se,$d,$i,$il) = (1,0,1,0,0);
- $s = 0 if (lc($x->{type}) eq "display");
- $d = 0 if (lc($x->{type}) eq "swish");
- ($s,$se,$d,$i) = (0,1,0,1) if (lc($x->{type}) eq "index");
- ($s,$se,$d,$i) = (0,1,0,0) if (lc($x->{type}) eq "swish_exact");
- ($s,$se,$d,$i,$il) = (0,1,0,0,1) if (lc($x->{type}) =~ /^lookup/);
+ my ($s,$se,$d,$i,$il) = init_visible_type($x->{type});
# what will separate last line from this one?
if ($display_data && $x->{append}) {
- $line_delimiter = ' ';
+ $line_delimiter = $delimiter;
} elsif ($display_data) {
$line_delimiter = '
';
}
@@ -228,7 +275,7 @@
if (($#data+1) == $nr) {
return sprintf($fmt,@data);
} else {
- print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n";
+ #print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n";
return $data;
}
} else {
@@ -257,7 +304,7 @@
$display = $new_display;
$cache->{lhash}->{$display} = $new_display;
} else {
- print STDERR "WARNING: lookup for '$display' didn't find anything.\n";
+# print STDERR "WARNING: lookup for '$display' didn't find anything.\n";
$display = "";
$cache->{lhash}->{$display} = $null;
}
@@ -278,48 +325,35 @@
}
# type="swish" ; field for swish
if ($swish) {
+ my $tmp = $swish;
if ($filter && ($s || $se)) {
no strict 'refs';
- my $tmp = join(" ",&$filter($swish)) if ($s || $se);
- $swish_data .= $tmp if ($s);
- if ($se) {
- if ($swish_exact_data) {
- $swish_exact_data .= "xxexx xxbxx ".$tmp;
- } else {
- $swish_exact_data .= $tmp;
- }
- }
-
- } else {
- $swish_data .= $swish if ($s);
- if ($se) {
- if ($swish_exact_data) {
- $swish_exact_data .= "xxexx xxbxx ".$swish;
- } else {
- $swish_exact_data .= $swish;
- }
- }
+ $tmp = join(" ",&$filter($tmp)) if ($s || $se);
}
+
+ $swish_data .= $tmp if ($s && $tmp);
+ $swish_exact_data .= "xxbxx $tmp xxexx " if ($tmp && $tmp ne "" && $se);
}
# type="display" ; field for display
if ($d && $display) {
+ my $ldel = $delimiter;
if ($line_delimiter && $display_data) {
- $display_data .= $line_delimiter;
+ $ldel = $line_delimiter;
}
if ($filter) {
no strict 'refs';
- if ($display_data) {
- $display_data .= $delimiter.join($delimiter,mkformat($x,&$filter($display)));
- } else {
- $display_data = join($delimiter,mkformat($x,&$filter($display)));
+ my @arr;
+ foreach my $tmp (&$filter($display)) {
+ my $tmp2 = mkformat($x,$tmp);
+ push @arr,$tmp2 if ($tmp2);
}
+ $display_data .= $ldel if ($display_data && @arr);
+ $display_data .= join($delimiter,@arr);
} else {
- if ($display_data) {
- $display_data .= $delimiter.mkformat($x,$display);
- } else {
- $display_data = mkformat($x,$display);
- }
+ $display_data .= $ldel if ($display_data);
+ my $tmp = mkformat($x,$display);
+ $display_data .= $tmp if ($tmp);
}
}
@@ -331,7 +365,7 @@
no strict 'refs';
$idisplay = &$filter($idisplay);
}
- push @index_data, $idisplay if (! $iterate_by_page);
+ push @index_data, $idisplay if ($idisplay && !$iterate_by_page);
}
# store fields in lookup
@@ -340,11 +374,21 @@
if ($lookup_key) {
print STDERR "WARNING: try to redefine lookup_key (keys shouldn't be repeatable fields!)";
} else {
- $lookup_key = $display;
+ if ($filter) {
+ no strict 'refs';
+ $lookup_key = &$filter($display);
+ } else {
+ $lookup_key = $display;
+ }
}
} elsif (lc($x->{type}) eq "lookup_val") {
if ($lookup_key) {
- $lhash{$lookup_key} = $display;
+ if ($filter) {
+ no strict 'refs';
+ $lhash{$lookup_key} = &$filter($display);
+ } else {
+ $lhash{$lookup_key} = $display;
+ }
} else {
print STDERR "WARNING: no lookup_key defined for '$display'?";
}
@@ -420,19 +464,17 @@
# now try to parse variables from configuration file
foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) {
+ $x = unroll_x($x);
+
my $delimiter = x($x->{delimiter}) || ' ';
my $val = $cfg->val($database, x($x->{content}));
- my ($s,$d,$i) = (1,1,0); # swish, display default
- $s = 0 if (lc($x->{type}) eq "display");
- $d = 0 if (lc($x->{type}) eq "swish");
- # no support for swish exact in config.
- # IMHO, it's useless
- ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
+ # FIXME index_lookup is not supported!
+ my ($s,$se,$d,$i,$il) = init_visible_type($x->{type});
if ($val) {
$display_data .= $delimiter.$val if ($d);
- $swish_data .= $val if ($s);
+ $swish_data .= " ".$val if ($s);
$index->insert($field, $val, $path) if ($i);
}
@@ -463,7 +505,8 @@
for (my $page=0; $page <= $nr_pages; $page++) {
my $display_data;
if ($cache->{format}->{$field}) {
- $display_data=mkformat($cache->{format}->{$field},$cache->{display_data}->{$field}->[$page]);
+ my $tmp = mkformat($cache->{format}->{$field},$cache->{display_data}->{$field}->[$page]);
+ $display_data=$tmp if ($tmp);
} else {
$display_data = $cache->{display_data}->{$field}->[$page];
}
@@ -484,7 +527,7 @@
$swish_data =~ s/ +/ /g;
$swish_data =~ s/ +$//g;
- $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
+ $xml .= xmlify($field."_swish", my_unac_string($codepage,$swish_data));
}
my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page];
@@ -494,7 +537,7 @@
# add delimiters before and after word.
# That is required to produce exact match
- $xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx'));
+ $xml .= xmlify($field."_swish_exact", my_unac_string($codepage,$swish_exact_data));
}
my $idel = $cache->{index_delimiter}->{$field};
@@ -527,7 +570,7 @@
$swish_data =~ s/ +/ /g;
$swish_data =~ s/ +$//g;
- $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
+ $xml .= xmlify($field."_swish", my_unac_string($codepage,$swish_data));
}
if ($swish_exact_data) {
@@ -536,7 +579,7 @@
# add delimiters before and after word.
# That is required to produce exact match
- $xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx'));
+ $xml .= xmlify($field."_swish_exact", my_unac_string($codepage,$swish_exact_data));
}
}
}
@@ -573,9 +616,18 @@
my $show_progress = $cfg_global->val('global', 'show_progress');
-my $unac_filter = $cfg_global->val('global', 'unac_filter');
-if ($unac_filter) {
- require $unac_filter;
+my $my_unac_filter = $cfg_global->val('global', 'my_unac_filter');
+if ($my_unac_filter) {
+ print STDERR "using $my_unac_filter to filter characters for search\n";
+ require $my_unac_filter;
+} else {
+ print STDERR "### fallback to default my_unac_string!\n";
+ eval q{
+ sub main::my_unac_string($$) {
+ my ($charset, $string) = (@_);
+ return $string;
+ }
+ };
}
foreach my $database ($cfg->Sections) {
@@ -589,6 +641,8 @@
#tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644;
tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644;
print STDERR "creating lookup file '$lookup_file'\n";
+ # delete memory cache for lookup file
+ delete $cache->{lhash};
}
# open existing lookup file
@@ -605,7 +659,7 @@
my $type_base = $type;
$type_base =~ s/_.+$//g;
- $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config', 'format' ], forcecontent => 1);
+ $config=XMLin("./import_xml/$type.xml", ForceArray => [ $type2tag{$type_base}, 'config', 'format' ], ForceContent => 1 );
# output current progress indicator
my $last_p = 0;
@@ -725,8 +779,9 @@
print STDERR "\n";
} elsif ($type_base eq "excel") {
- use Spreadsheet::ParseExcel;
- use Spreadsheet::ParseExcel::Utility qw(int2col);
+ require Spreadsheet::ParseExcel;
+ require Spreadsheet::ParseExcel::Utility;
+ import Spreadsheet::ParseExcel::Utility qw(int2col);
$import2cp = Text::Iconv->new($config->{excel_codepage},$codepage);
my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!";
@@ -780,29 +835,32 @@
}
} elsif ($type_base eq "marc") {
- use MARC;
+ require MARC;
$import2cp = Text::Iconv->new($config->{marc_codepage},$codepage);
my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!";
# optional argument is format
- my $format = x($config->{format}) || 'usmarc';
+ my $format = x($config->{marc_format}) || 'usmarc';
print STDERR "Reading MARC file '$marc_file'\n";
my $marc = new MARC;
my $nr = $marc->openmarc({
file=>$marc_file, format=>$format
- }) || die "Can't open MARC file '$marc_file'";
+ }) || die "Can't open MARC file '$marc_file' with format '$format'";
+
+ # read MARC file in memory
+ $marc->nextmarc(-1);
- my $i=0; # record nr.
+ my $max_rec = $marc->marc_count();
- my $rec;
+ for(my $i=1; $i<=$max_rec; $i++) {
- while ($marc->nextmarc(1)) {
+ progress($i,$max_rec);
- # XXX
- fakeprogress($i++);
+ # store value for marc_sf.pm
+ $main::cache->{marc_record} = $i;
my $swishpath = $database."#".$i;
@@ -814,6 +872,9 @@
print "Document-Type: XML\n\n$xml\n";
}
}
+
+ print STDERR "\n";
+
} elsif ($type_base eq "feed") {
$import2cp = Text::Iconv->new($config->{feed_codepage},$codepage);
@@ -868,6 +929,10 @@
all2xml.pl - read various file formats and dump XML for SWISH-E
+=head1 SYNOPSYS
+
+ $ all2xml.pl [test.conf]
+
=head1 DESCRIPTION
This command will read ISIS data file using OpenIsis perl module, MARC
@@ -876,6 +941,9 @@
this script B from isis files (isis allready
has something like that). Output of this script is tailor-made for SWISH-E.
+If no configuration file is specified, it will use default one called
+C.
+
=head1 BUGS
Documentation is really lacking. However, in true Open Source spirit, source